J . D U R K A N a n d R . K E L L E H E R
S
INCE S e p t e m b e r 1968 T h e E c o n o m i c a n d Social Research I n s t i t u t e has p u b l i s h e d d e t a i l e d forecasts o n a q u a r t e r l y basis o f Gross N a t i o n a l P r o d u c t a n d its c o m p o n e n t s i n b o t h v a l u e a n d v o l u m e t e r m s . I t is h o p e d i n tins p a p e r t h a t b y e x a m i n i n g relationships b e t w e e n forecasts a n d t h e o u t c o m e s , o n e m a y be able t o g e t s o m e i n s i g h t s i n t o t h e adequacy o f f o r e c a s t i n g m e t h o d s e m p l o y e d i n this c o u n t r y . W e w i l l also a t t e m p t t o take cognisance o f q u a l i t a t i v e statements c o n t a i n e d i n t h e v a r i o u s p u b l i c a t i o n s c o n t a i n i n g forecasts.S o m e o b j e c t i o n s m a y be raised t o t h e m e t h o d w e h a v e e m p l o y e d i n this p a p e r t o test t h e f o r e c a s t i n g r e c o r d . I n p a r t i c u l a r , i t c o u l d be a r g u e d t h a t b y m e r e l y t a k i n g t h e figures o u t o f t h e v a r i o u s c o m m e n t a r i e s w i t h o u t p a y i n g a n y a t t e n t i o n t o t h e n u m e r o u s reservations a n d q u a l i f i c a t i o n s t h a t w e r e expressed a b o u t those figures i n t h e a c c o m p a n y i n g texts w e h a v e n o t d o n e f u l l j u s t i c e t o t h e forecasters' v i e w s o f trends i n t h e e c o n o m y . Several p o i n t s need t o be m a d e i n this r e g a r d . First, t h e figures d o , i n fact, represent w h a t o n balance t h e forecasters felt w a s l i k e l y t o o c c u r . I t is t h e figures, a n d n o t t h e q u a l i f i c a t i o n s , w h i c h r e c e i v e d p u b l i c i t y a n d , thus, these figures e m b o d y w h a t e v e r i n f l u e n c e o n decisions such c o m m e n t a r i e s possess. T h e r e f o r e a test o f t h e accuracy o f t h e figures f o r m s t h e essential c o n t e x t w i t h i n w h i c h a n y j u d g m e n t o f t h e forecasts m u s t be m a d e . S e c o n d , w e feel o u r analysis i m p l i e s t h a t m o r e emphasis s h o u l d be p l a c e d o n t e x t u a l a n d q u a l i t a t i v e analysis o f trends i n t h e e c o n o m y r a t h e r t h a n t h e precise q u a n t i f i c a t i o n o f f u t u r e events. A t best t h e present degree o f accuracy o f f o r e casting m e t h o d s is t h a t statements such as " t h a t g r o w t h is l i k e l y t o be a b o v e ( o r b e l o w ) a v e r a g e " c a n be m a d e . T h u s , t h e fact t h a t t h e absolute levels o f past forecasts h a v e i n s o m e cases d i v e r g e d q u i t e a l o t f r o m t h e e v e n t u a l o u t c o m e s does n o t i m p l y t h a t t h e c o m m e n t a r i e s themselves h a v e n o t been v a l u a b l e i n t h e
understanding o f current trends in the economy. Unrealistic expectations have been generated about the accuracy o f the forecasts and it is important that the large confidence intervals attached to the forecasts be stressed.
Divergences between forecasts and eventual outcomes do not necessarily reflect faults in the forecasting-, process. Any set o f forecasts are based on certain assump tions and to the extent that these assumptions are unrealised, forecasts can be expected to differ from outcomes. Indeed, in some cases the forecasts themselves could result i n changes in some o f the assumptions on which they were originally based. I f sluggish growth i n the economy were forecast on the basis o f unchanged fiscal and monetary policies, the effect o f such a forecast could well be that the' authorities would take a considerably more expansionary stance on these issues which would thus revise upwards the expected outturn of economic activity in that year.
Apart from these "policy" assumptions many others have to be made in relation to exogenous influences on the economy which will largely be determined by political and other non-economic factors in which the forecaster has no particular expertise. Present forecasts have to make some assumptions concerning the likely future development o f oil prices. To the extent that eventual divergences between forecasts and outcomes can be accounted for by the failure o f these assumptions to be realised no blame can be attached to the forecasting process.
The Forecasting Record
These reservations aside, a systematic divergence between forecasts and out comes may be due to inadequacy in forecasting procedures. W i t h the possible exception o f the "policy" assumptions the effect o f the non-fulfilment o f initial expectations concerning other exogenous influences should not bias the forecasts in any particular direction all the time. It is possible that both fiscal and monetary policy has tended to be rather more expansionary usually than was originally assumed, so that some downward bias in the forecasts would be expected. To some extent, however, this problem can be overcome by looking at post-budget forecasts to see to what extent they differ from previous ones. I t is our contention that a systematic underestimation has occurred in the forecasts. This trend seems to be considerably stronger in value forecasts than those relating to volume suggesting that the increase i n price deflators has been subject to considerable underestimation.
Principal conclusions appear obvious from inspection o f Tables I to 3 which present the various forecasts made for each year together with three different estimates o f the eventual outcomes for those years. In these Tables five forecasts are given for each year. The first o f these is normally made in the autumn o f the previous year with further forecasts coming i n the following winter, spring, summer and autumn. A l l forecasts relate to expenditure on Gross National Product and no attempt has been made to assess forecasts made on the incomes and output sides.
QEC QEC QEC . QEC QEC First CSO CSO 1st 2nd :3rd 4th sth Official 1st . Latest Forecast Forecast Forecast Forecast Forecast Estimate Estimate Estimate
September January May September December
J 969 1968 1969 1969 1969 1969
c 11-5 7'5 10-3 12-5 12-0 n - 5 12-0 13-5 G 8-5 8/5 . 10-4 10-5 I I ' O 13-0 16-25 16-1 . I 17-0 15-5 20-8 22'0 22-0 27-0 3i-o 29-0
X 6-0 8-6 I O ' O io-o 12-0 . 9-5 n - o •
M I 2 ' 0 8-o ' 15-0 18-5 17-5 18-0 r8-o 18-2 " G N P 9-8 8-5. ,. 10-5 n - 5 10-5 12-0 12-25 14-5
September December March June September
.
1970 1969 1969 . 1970 1970 1970
C 9-0 I I ' O • I I ' O n - 5 io-s 9-75 to-75 10-75
G I O ' O 12-0 13-0 13-0 12-5 18-75 20-5 20-5
I I I / O 14-5 15-0. 13-0 12-5 7-25 7-0 8-1
X n - 5 12-0 12-0 I2'0 I 2 ' 0 13-75' 12-5 n - 7 M 10-5 n - 5 12-5 9-5 9-0 10-25 10-25 9-5 G N P n - o n - 5 11-5 12-0 n - 5 11-25 n - 5 n - 7
September December March June Autumn
1971 1970 1970 1971 1971 1971
C 9-0 9-5; io-o I I ' O I 2 ' 0 11-25 12-5 10-9 G 12-0 10-5 10-5 n - 5 13-5 17-25 ' 17-5 22-3 I 15-0 14-0 14-0 17-0 15-0 17-5 21-0 19-8 X n - 5 I2'0 12-0 12-5 12-0 I I - 5 . 12-5 12-6 M 14-5 13-5 13-0 14-5 13-0 10-25 10-5 . 10-3 G N P 9'5 I O ' O 10-5 I I ' O n - 7 5 12-75 14-0 13-3 '
Autumn Winter July October January
1972 1971 1971/72 1972 1972 1973
C 10-5 7-75 I I ' O 11-25 12-0 12-75 13-9 13-9 G 12-5 14-0 1 14*0 l 6 -5 17-0 22-5 25-4 25-4
I 9-0 5*5 • 8-5 8-5 8-5 n - 5 12-5 12-5 X 12-5 .8-0' 9-0 9-5 14-5 15-5 14-25 14-25 M n - 5 7*5 7-5 7-0 10-5 11-25 11-25 11-25 G N P 11-0 9-0 11-5 12-5 13-5 16-5 i8-o 18-0
October January • April August December
1973 1972 1973 1973 1973 1973
C 10-5 I I - S 13-0 15-75 18-75 18-4 G 14-5 15-0 17-0 20*0 21-0 20*7 I 17-0 17-0 19-0 24'0 27-0 27-6 X 15-0 20-5 ' 22-5 28-0 31-25 31-4 M 14-0 17:0 21-s 29-5 34-o 34-2 G N P 13-25 14-0 12-0 18-0 20-75 19-5
forecasts relating to current values. I n all the years considered the growd* in the value o f GNP has been greater than suggested by the forecasts. This divergence has become greater in the last two years.For 1972 the first forecast was for a rise of 11 per cent in the value o f Gross National Product while the latest estimate is for an 18 per cent rise. The first forecast for 1973 was for a 13 per cent rise while the fifth forecast was over 20 per cent. This pattern is true for practically all o f the components of GNP. However, most surprisingly, the greatest error made, both in forecasting and estimating, was in the growth o f government expenditure. In all years the latest estimate for the percentage increase in this item is approximately twice the original forecast. Furthermore, this cannot be explained by fiscal policy being more expansionary dian was originally assumed for even the final forecasts for each year considerably underestimate the true growth in public expenditure. O f course, in 1973 the greatest error was in relation to the. growth o f both exports and imports. The final forecasts for that year were for a 31*25 per cent and 34 per cent increase respectively while the first forecasts were 15 per cent and 14 per cent. No such clearcut conclusions emerge from Table 2 which presents volume forecasts for GNP and its components. Probably the most important forecast made is that for the growth i n the volume o f GNP. This indicator is best accepted as showing the degree o f under-utilisation or inflationary pressure which exists in the economy. Unfortunately, the record o f forecasting these growth rates is not altogether too impressive. The picture clearly emerging from examination o f the forecasts is that they have tended to cluster around a mean range o f 3-5 per cent. Thus, in years o f reasonably average growth, such as 1971 and 1972, the forecasts have been fairly accurate. However, when the growth rate has diverged from the average the forecasts have been unable to detect the extent o f this divergence. 1968 and 1973 were years o f very much above average growth. The estimated growth rate (on expenditure data) for 1968 is now 8-5 per cent while the corresponding preliminary estimate for 1973 is 6-9 per cent. The initial forecast for 1968 was 3-6 per cent while the first post-budget forecast in May o f that year was 3*9 per cent. The initial forecast for 1973 was 5-25 per cent with an immediate post-budget expectation o f 5-75 per cent. O n the other hand, 1970 was a year o f sluggish growth with GNP rising in volume terms by 2*7 per cent. The first prediction was for an increase o f 5 per cent which represented a higher growth than at the time was estimated for 1969. Little can be said about the value of forecasts in this year since major distortions were introduced by particularly bad industrial relations. The general conclusion is that the forecasts have not been successful in identifying the extent to which the economy sometimes diverges from its average growth path. A t best all that the forecasts are likely to be able to say is either that growth w i l l be above, below or about average in the coming year or perhaps above or below the previous year's level. I t is important that quantita tive forecasts presented should be interpreted in this way.
hope-TABLE 2 : Constant Values (per cent change)
QEC QEC QEC Q E C QEC First CSO CSO
1st 2nd ird 4th 5th Official 1st Latest
Forecast Forecast Forecast Forecast Forecast Estimate Estimate Estimate
1969
c 5-3 2-9 3-6 4*5 4-0 4-0 4-5 5'4 G 2-4 2-4 2-6 2-5 3-0 4-0 8-5 6-9 I 10-5 8-9 13-0 13-0 13-0 17-5 23-25 19-7 X 3-5 4-4 6-0 6-0 5-5 6-o 3-5 4'7 M 8-8 4-6 n - 7 13-0 12-0 n - 5 13-0 13-5 G N P 4-2 4-0 3-8 4-0 4-0 3-75 4-25 5-3
1970
C 3-5 5-0 5-0 3-0 2-5 1-5 2-75 2-1 G 3-0 2-5 3-0 3-o 3-0 3-75 7-0 6-3 I 9'S 6-0 6-5 3-5 3-5 - 1 - 2 5 - P 2 5 — X 8-0 8-0 8-5 8-0 4'5 6-0 5-25 4-7 M 7-0 8-5 9-0 6-0 2-0 3-5 3-25 2-5 G N P 5-0 4'5 4-5 3-0 2-5 i-5 2-75 2-5
1971
C 3-0 2-5 2-5 3-o 3-0 2-0 3-25 2-1 G 3-5 2-0 2-0 2-0 3-0 6-25 3-25 11-25 I 8-o 6-0 8-0 9'5 7-5 8-o 9-0 9-25 X 6-0 6-0 6-0 5-5 5-0 3-5 5-0 5-0 M 9-0 7-0 7-5 8-0 6-0 3-0 4-0 4-0 G N P 3'5 2-75 3-0 3-0 3-0 3-o 3-0 2-9
1972
C 3-5 0-5 2-75 2-5 3-0 4-0 5'4 5*4 G 3-0 4*5 4-5 5-0 5-5 8-75 10-4 10-4 I 2-5 - o - 5 2-0 2-0 2-0 2-0 2-5 2-5 X 6-0 2-5 2-5 2'0 2'25 3'5 2-0 2-0 M 6-0 i-5 2-5 2-0 5-0 7-25 7*9 7-7 G N P 3'75 1*5 3-0 2-75 2-75 3-0 3-7 4-2
1973
C 3-5 3-5 4-25 5-0 7-0 6-3 G 5-0 5-0 6-0 8-0 8-0 7-4 I io-o io-o io-o 12-5 15-0 12-9 X 8-0 12-0 14-0 1 5 7 5 16-75 io-6 M 8-5 I P O 14-0 19-5 22-25 13-5 G N P 5-25 5-25 5-0 5-75 7-0 6-6
lessly u n d e r e s t i m a t e d the g r o w t h i n the v o l u m e o f c u r r e n t ' g o v e r n m e n t c o n s u m p t i o n e x p e n d i t u r e . I n 1971 the e x p e c t a t i o n was f o r a 3 p e r cent g r o w t h w h i l e the actual t u r n e d o u t t o b e o v e r 11 p e r cent. A s i m i l a r e r r o r o c c u r r e d i n 1972.
A s f o r t h e o t h e r c o m p o n e n t s m a n y o f t h e c o m m e n t s m a d e i n r e l a t i o n t o G N P are a g a i n a p p l i c a b l e . I n years o f a b o v e average g r o w t h (1968 a n d 1973) the forecasts u n d e r e s t i m a t e d t h e e x p a n s i o n i n b o t h the e x t e r n a l sector a n d i n c a p i t a l f o r m a t i o n a l t h o u g h a considerable u n d e r e s t i m a t i o n o f c o n s u m e r e x p e n d i t u r e also o c c u r r e d i n 1973. I n 1970, a y e a r o f v e r y s l u g g i s h g r o w t h , considerable o v e r e s t i m a t i o n i n the g r o w t h o f all c o m p o n e n t s o f G N P o c c u r r e d .
These characteristics o f t h e forecasts are u n l i k e l y t o d e p e n d o n the i n s t i t u t i o n m a k i n g t h e forecasts. R a t h e r , t h e y reflect the general state o f t h e a r t o f f o r e casting i n t h e c o u n t r y i n general o r perhaps e v e n the w o r l d w i d e state o f the art. Forecasts* w e r e m a d e f o r years earlier t h a n 1969 i.e. before t h e " Q u a r t e r l y E c o n o m i c C o m m e n t a r y " Series c o m m e n c e d . I n 1965 a n d 1966 t h e D e p a r t m e n t o f Finance forecasts w e r e u n a b l e t o p i c k u p the e x t e n t o f the d o w n t u r n o f the e c o n o m y i n those years. I n 1965 t h e i r e x p e c t a t i o n was 4 p e r c e n t g r o w t h , w h i l e t h e o u t c o m e t u r n e d o u t t o b e 2-75 p e r cent. I n 1966 t h e i r " p e s s i m i s t i c " o u t l o o k was f o r a 3*7 p e r c e n t g r o w t h rate w i t h the o u t c o m e e v e n t u a l l y b e i n g 1-25 p e r cent. A g a i n i n 1967 the N I E C w e r e u n a b l e t o i d e n t i f y the e x t e n t o f the r e c o v e r y i n t h e e c o n o m y . I n recent years d i e E S R I has been t h e o n l y b o d y t o p u b l i s h d e t a i l e d forecasts a n d t h e y c o n t i n u e , as has been discussed, t o h a v e s i m i l a r characteristics. I t m u s t b e assumed t h a t these characteristics h a v e also c o n t i n u e d t o b e p a r t o f t h e forecasts o f o t h e r I n s t i t u t i o n s .
T a b l e 3 gives the v a r i o u s forecasts f o r the increase i n the p r i c e deflators o f G N P a n d its c o m p o n e n t s . T h e u n d e r e s t i m a t i o n w h i c h o c c u r r e d i n the v a l u e forecasts was m a i n l y d u e t o a v e r y s i g n i f i c a n t u n d e r e s t i m a t i o n o f the increase i n p r i c e levels. E x c e p t f o r i m p o r t prices i n 1972 the i m p l i c i t p r i c e deflators o f G N P a n d its c o m p o n e n t s h a v e b e e n u n d e r e s t i m a t e d i n e v e r y year since 1969. T h e result has been t h a t e v e n i n some cases w h e r e v o l u m e increases w e r e c o n s i d e r a b l y o v e r e s t i m a t e d the u n d e r e s t i m a t i o n o f d i e p r i c e increase resulted i n the o v e r a l l u n d e r e s t i m a t i o n o f v a l u e forecasts, (e.g. e x p o r t s i n 1972).
T h e p r i c e forecast o n w h i c h m o s t interest centres is t h a t f o r c o n s u m e r prices. A g a i n , w e h a v e c l e a r l y b e e n .unable t o grasp d i e e x t e n t t o w h i c h c o n s u m e r prices w e r e increasing i n a n y year u n t i l the f i n a l p a r t o f the year. T h u s , i n e v e r y year the forecast was c o n t i n u a l l y revised u p w a r d s f r o m c o m m e n t a r y t o c o m m e n t a r y . F o r e x a m p l e , t h e i n i t i a l e x p e c t a t i o n f o r 1973 was a rise o f 7 per c e n t — a n e x p e c t a t i o n c o n t i n u a l l y revised u p w a r d s t o t h e f i n a l f i g u r e o f 11*4 p e r cent. Possibly, d i e forecasts represented a h o p e (rather t h a n a b e l i e f ) d i a t successful a n t i - i n f l a t i o n a r y policies w e r e b e i n g i n t r o d u c e d . F u r t h e r m o r e , i t is possible t h a t forecasts o f v e r y
TABLE 3: Price Deflators (per cent change)
QEC QEC QEC QEC QEC 1st cso CSO
1st 2nd 3rd 4th Sth Official 1st 1st
Forecast Forecast Forecast Forecast Forecast Estimate Estimate Estimate
1969
c 6-o 4-5 6-5 7-5 rs 7-25 7-2S 8-0 G 6-o 6-o 7'5 7*5 7*5 8-75 7-25 8-5 I 6-0 6-o 7-0 8-0 8-0 8-0 v6-25
7-75 X i-5 i-5 2-5 4-0 4-0 5-75 5'75 6-0 M 4-75 8-0 3-0 5-0 5-0 5-75 4-5 4-1 G N P 5-5 4-3 6-5 7-0 7-0 8-0 7-75 8-7
1970 •
C 5-5 5-5 6-0 8-5 8-0 8-0 7'75 8-5
G 7-0 9-0 I O - O I O - O 9-0 9-5 12-5 13-5
I 7-0 8-o 8-0 9-5 8-5 8-5 8-5 8-0 • X 3-0 3'5 3-5 4-0 7-0 8-0 7-0 6-75 M 3-0 . 3-0 3-0 3-5 6-5 7-5 6-75 6-75 G N P &o 6-5 6-5 9-0 8-5 8-5 8-5 9-0
1971
C 6-o 7-0 7-0 8-o 8-5 9-0 8-5 8-5 G 8-0 8-5 8-5 9-0 io-o 10-5 13-75 io-o I 6-5 7-5 6-0 7-0 7-0 8-75 I I - O 9'5 X 5-0 6-o 6-0 6-5 6-5 7'75 7-0 7'5 M 5-0 6-0 5-0 6-o 6-s 7-0 6-25 6-20 G N P 6-0 7'5 7-5 8-o 8-25 9-5 10-75 i o - i
1972
C 7-0 7-25 8-0 8-5 8-5 8-5 8-1 8-1
G 9-0 9-0 9-0 I I - O i i - o 12-5 13-6 13-6
I 6-0 6-o 6-s 6-5 6-s 9-25 9-8 9-8 X 6-0 5-5 6-s 7-5 12-0 i i - 5 12-0 12-0 M 5*5 6-o S-o S-o 5-0 6-5 3-2 3-2
G N P 7-25 7-25 8-25 9-25 io-5 13-0 13-6 13-2
1973
C 7-0 7-5 8-5 10-25 I I - O G 9-0 9-5 io-o I I - O 11-75 I ' 6-5 6-5 8-0 io-o 10-5 X 6-5 7-5 8-0 10-75 12-25 M 5-o 5'5 6-5 8-75 9-5 G N P 7'5 8-5 9'5 " • 5 12-75
rapid growth o f consumer prices could in themselves have inflationary effects— a situation which the Institute would be anxious to avoid. Hopefully, our present forecast o f a rise o f 15 per cent for 1974 in consumer prices is not subject to similar large errors as in the past!
Government expenditure in Table 3 reflects again the findings of Tables 1 and 2 in that it seems quite often to have the largest margin o f error. The other components have similar trends as that for consumer prices—an initial under estimation which is continually revised upwards throughout the year.
The statistical significance o f the divergences discussed .above was tested by regressing the outcomes on the forecasts. Ideally, one would wish to do this for each component o f GNP. However, since observations on both forecasts and outcomes exist for only four years this was clearly ruled out. Instead, all forecasts for current values were considered as one variable and, similarly, for constant values and price deflators. This resulted in 24 observations on each variable when forecasts were paired with first official estimates (FO) or CSO first estimates and 18 observations when paired with CSO 1972 observations. However, any results obtained must be related to a group o f forecasts and no inferences can be drawn for the individual components o f GNP.
Since practically all the equations had a constant term not significantly different from zero they were re-estimated through the origin. I f the resulting coefficient was significantly different from unity we concluded that the forecasts were significantly different from the outcomes.
The results for current and constant values are given in Tables 4 and 5 respec tively. This confirms that systematic underestimation occurs as practically all o f the coefficients are significantly greater than unity. Furthermore, the coefficients decline, as one moves from first forecasts to final forecasts, confirming our belief of a constant upward revision throughout the year. In general, the degree o f underestimation appears to be greater in relation to the CSO estimate than to the FO estimates, reflecting that these latter estimates also tended towards under statement.
Table 5 again seems to confirm our expectations. On average, forecasts o f volume changes in the economy have not been significantly different from the outcomes. However, the regression analysis does not bring out the extent to which the forecasts have failed whenever the economy diverges from average values. Interestingly, final forecasts again tend to understate significantly the CSO estimates but not FO estimates. The overall effect is to re-emphasise the understatement o f the increase in general price levels.
Problems of Data
TABLE 4 : Results: Current Values
Observations 1969-1972 (inclusive) t(B — 1)
DW
F O = 1-25 FX
( i 8 -3)
F O = 8 - 4 i + - 5 4 Fs
(2-45) (1-67) F O = 1-18 F ,
(19-17) F O = I - I I F4
(21-93) F O = I - I I FS
(28-75)
C S / = 1-3 3 FX
(16-89) C S I = 1-3 8 F2
(I3-79)
CSI = I - 2 5 F ,
(17-79)
C S I = I - I 8 F4
(20-36)
C S I = I - I 8 F6
(23-77) C S I = I - 0 7 D . F .
(56 - i )
3- 68**
t
2-88**
2-18*
2- 84**
4- 21**
3- 8 3 * *
3'59**
3-17**
3-62**
3.48**
1-79
1-97
1.82
1- 86
2- 43
1-76
1-27
1-65
1- 70
2- I I
1-78
•53
•34
•58
•70
•83
•59
•23
•64
•73 •81
•97
Observations 1969-1971 (inclusive) t(B = 1) DW R
CSL = I-32FJ. 3-29** 1-59 •52
(13-7)
•52
CSL = 9-47 + -47FSS t 1-85 • -23
(i-66) (-93)
• -23
CSL = I T 9 F 3 2-26* I - 8 I •56
(14-12)
•56
CSL = i - i 3 F4 1-85 2-23 •68
(16-16)
C S L = i - 6 7 F6 2-65* 2-22 •77
(18-55)
•77
CSL = i - o 8 D . F . 3-13** 1-59 •96
(13-36)
•96
CSL = i-oi C S I •38 2-46 •95
(38-43)
•95
See note Table 1.
E C O N O M I C A N D SOCIAL REVIEW
TABLE 5: Constant Values: Results
Observations 1969-1972 (inclusive) t(B= 1) DW R
F O = - 85F - 1 - 1 8 1-53 -37
(6-92)
F O = 2 - 8 o + -5F8 t 1-67 -33
(1-91) (1-62)
F O = - 8 9 F , —1-04 1-46 -59 (8-45)
F O = - 9 5 F4 - - S i 1-52 -73
(10-38)
F O = i - i o F6 1-37 1-91 -86
(I4-75)
•32
•32
'57
•69
•84
•93
CSL = - 9 8 F ! —-12 1-47 -29
(5'7)
CSL = 2-8 + - 5 F * t 1-69 -33
^•91) (1-62)
CSL = i - 9 7 Fs —-23 i-5s .50
(6-57)
CSL = I - C H F4 -32 i-8o -67
(8-04)
CSL = i - 2 5 F6 2-17* 2-12 -82
(10-92)
CSL = I - I 7 F 2-91** 1-87 -94
(19-68)
CSL = - 9 4 C S / —-90 -co -89
(13-98)
CSI = i-ooFl o-o 1-63
(6-28)
1-63
CSI = 3 - i 8 + - 6 i F2 t 1-87
( i - 7 i ) ( i ' 5 6 )
1-87
CSI = i - 0 5 F8 •40 1-58
(7-76)
C S r = I - I 3 F . 1-04 1-56 (9-22)
CSI = i - 3 i F6 3-04 1-83
(12-76)
CSI=viSD.F. 2-98** 1-87
(19-89)
1-87
up to the third quarter (and again the third quarter figure has only recently become available), no data at all on wholesale prices and no data on agricultural output except for the livestock enumeration. These delays mean that commentators very often have to "forecast" the past before making projections into the future. Clearly, forecasts will reflect the cumulative errors o f such a procedure. Until these delays are greatly reduced the accuracy of forecasts is not likely to be radically improved.
The second dimension is that data available to forecasters is preliminary and sometimes subject to drastic revisions. T w o consequences follow. First, i t distorts one's view o f the past relationship between variables in the economy. Whenever econometric methods are being used the coefficients o f estimated equations w i l l be distorted depending on the extent to which preliminary rather than final data was used i n their estimation. Denton and Kuiper [ 1 ] estimated simple macro-models for several countries using preliminary, mixed and final data. They found the resulting coefficients highly sensitive to the type o f data used in estimation and much less sensitive to different methods o f estimation. This also holds when non-econometric methods (such as mere inspection) are being used.
Second, data revisions also distort one's view of current trends in the economy. It w i l l be argued that such a distortion was in large part responsible for the considerable underestimation of the growth in the economy in 1973. Denton and Oksanen [ 2 ] examined preliminary and final data for fourteen countries and found that, in general, preliminary data is later subject to upward revision. They then estimated a simple macro-model for each country using only final data, (i.e. ignoring the first problem o f data revisions). Ex-post forecasts were generated feeding both preliminary and final data into the models. Forecasts generated by the preliminary data (which itself was an underestimation o f the final data), yielded forecasts significantly lower than those generated by final data. Clearly, this tendency holds also for forecast based on Irish preliminary data.
C s rt - v-i r q Q 7* " 7* " 9 ^ Co 6 N N t ^s V M\QO O O N « r$ w v-i M \ b W < S M M M M O ^ H t r ) C \ O N C O c o s o c o c o O N O A N O c o M d I H M M M w N 0 \ 0 \ h-t 0 \ rt) H 4 t-, V V D O O c o O M r l c o s p r} r< l o o M C** , K H 6 N t-4 M M t H W W O s c o O 'A H v % oo b b\ b\ v o 0 \ v o t> -M c o O t > V O V t rq s b 6 N N 7 * 7* - O N t ^o o p oo m fs' O ' O O N CO O O c o O O c o c o oo r o O v-i <-n H c O w " O c o M N v-» C N ^ * V p o o -V C O t > S O O S V \ M O O «~ o V O ^1 * O S !> t > «/ o c o C o S O 0 \ r o Co t^ . d o f» s o 0 0 0x0* 0 M ^ - O O N O O O O O t> o b d o Co Co c b C\c o o o c o -t C - o \ o \ s o v o Co c* ^ C O s o 0 0 O r o S O N O N M O i N VO c O C O C O -i-j C o so o \ c o \ p CO ^f-V Q M V CO I S < A H V O P O CO v o t > c o c o s o co s o O c o v o C \ O O t > r o v o N N c o 6 O c o p \ c o C O c o O C o v o b \ c o C o 0 0 ^4 * C O C O N 0 \ H N Co c 3 Co TJ -(S v o c o O * 0 t* -C O C O C O * H O C O vo v o o c o s b C \ H M O C O M O W Tt -t-t H i r o 6 C O M
I H
official estimates (FO)
C G I X M GNP
Current Values
1964 o-8 2-3 - 2 - 7 i ' 4 0-4 - o - 3
1965 1-5 3-9 1-6 — I T 0-4 O'O
1966 2-5 - 1 - 8 I T 2-6 2-8 1-8 1967 I T o-8 I ' 2 - 2 - 7 - 1 - 8 i-3 1968 3-9 - 1 - 5 - 0 - 6 —0*2 0-9 3-1
1969 2-0 3-1 2-0 — I T 0-4 2-4
1970 0-9 1-7 o-6 — 2 T - 1 - 8 0-4 1971 - o - 3 5-0 2-2 I ' 2 O T 0-5 Average I T 8 2'06 0-53 - • 2 5 0*2 I T 5 /Average/* 1-63 2-51 1-6 I T 1-22
Constant Values
1964 1-2 — 2 T — 2 T 3'2 2-5 —o*5 1965 — I ' O 1-7 0-5 0-4 0'2 0-3 1966 i-5 —0-2 2'0 1-7 3*5 — O T 1967 1-4 2-3 0-9 - 2 - 4 — 2 T 1-5 1968 3-2 2-5 — I ' O 0'2 I ' 2 3-1
1969 1-5 2-9 2 T —1-4 2 T 1*5
1970 o-6 2-5 1-3 - i - 3 — I ' O I ' O 1971 0-0 4*9 O'O i-3 I ' 2 — O T Average 1-05 I - 8 I •53 • 2 1 •95 •84 / A v e r a g e / * i-49 2-39 i - 4 i 1-49 1-73 I ' O I
Price Defl, ators
1964 — O T 4*7 - 0 - 4 - 1 - 9 —2*0 0-3 1965 - o - 4 2 T 0-9 —1*5 0'2 - 0 - 3 1966 0-9 - 1 - 5 — I ' O 0-8 —0-7 1-8 1967 - 0 - 3 - 1 - 5 0-3 - o - 3 0-3 —0-2 1968 0-5 — I ' O - o - 5 - 0 - 4 - o - 3 — O T 1969 - 0 - 4 — I ' O —0'2 0'3 - i - 6 0-7 1970 0 ' 2 — I T —0-8 - 0 - 7 0-2 - 0 - 7 1971 0-3 - o - 3 —2-0 0 ' 2 — I T o-6 Average. O T o r •46 - 4 8 - 6 3 •26 / A v e r a g e / * ,0-4 1-7 •76 •76 •80 •60
See note T a b l e 1 .
TABLE 8: Regression of First Official Estimates (FO) on CSO Latest Estimates (CSL)— (igSf-itfi)
Current Values r(B = i )
All variables: FO = -gi CSL —4-04
C FO — -87 CSL - 2 - 5 2 G FO = -98 CSL - 1 - 3 2 I FO = -97 CSL - - 8 i
X FO = I - O I CSL -09
M F O = -97 CSL - -72
G N P F O = -90 CSL - 2 - 7 7
Constant Values
All variables: FO = -88 C S L . —3-90
C F O = -68 C S L - 5 - 3 0 G F O = -6o C S L - 5 - 4 4 I F O = -99 C S L - 0 - 2 4 X F O = -96 C S L - o - 4 M F O = -86 C S L - 2 - 1 8
GNP . F O = -74 C S L - 4 - 4
Price Deflators
All variables: FO = -98 C S L . — -89
C F O = -98 C S L - -6o G F O = -93 C S L - -91
I FO = -97 C S L - -62
X F O = 1-08 C S L 1-34
M F O = I - I O C S L 1-27 GNP FO = -96 CSL - -98
See note Table 1.
mainly due to the underestimation o f private and government consumption expenditure. Thus, the quality o f data and forecasts are inextricably related.
was due to preliminary data understating the final outcomes. However, the fact that the greatest underestimation occurred in forecasts of price deflators cannot be explained in this way. Table 7 shows that, in general, revisions o f price deflators have been small and to a large extent appear to be random. This is borne out by Table 8 where more of the coefficients for price deflators are significantly different from unity—no matter whether all variables are taken together or individual components are tested separately.
The Forecasts for 1973
Apart from these general observations each year has its own peculiarities. I t is worthwhile to look at some o f the reasons for the changes made i n the forecasts during 1973. The forecasts o f the Institute are built up primarily around an expenditure table for GNP in current prices, with a supporting table for National Income which acts as an input into the derivation o f the expenditure table and provides a consistency test o f the orders o f magnitude involved in that table.
The growth rate, i n real terms, derives, given the GNP figure, from the price indices applied to the components o f the expenditure table. Changes i n the forecast growth rate result both from changes in value and in the price indices used. To date the Institute has published five forecasts for the year 1973. Even at this stage it would be very difficult to go much beyond the statement made in the first forecast in October 1972 that 1973 would be a year o f very rapid growth. It is possible, however, to pad out the picture, on the expenditure side w i t h figures in current prices for private consumption, the external sector, and to a much lesser degree government consumption, and investment. The forecast in value terms in the December commentary is more than likely reasonably correct— requiring revision in the light o f the figures in N I E (1972). In value terms GNP is likely to have risen by about 2 1 per cent over 1972, compared with the first forecast o f 13*25 per cent in October 1972. The table below indicates the change in the pattern o f the forecast in the five issues o f the commentary concerned with
1973'-Percentage change over 1972
October January April August December
Commentary 1972 1973 1973 1973 1973
G N P current prices 13-25 15 18 20-75 G N P constant prices 5-25 5-25 5 5*75 7 G N P deflator 7-50 8-50 9-50 11-50 12-75
is roughly equivalent to the earlier forecasts when allowance is made for this. While the table is summary in nature it does reveal a failure to forecast adequately both price and volume in 1973.
A t this time (May 1974) we would be reluctant to say what the growth rate for the year was. This reluctance stems primarily front lack o f data on prices in general so that the appropriate price indices cannot be applied to the components of GNP presented in the December commentary. This lack is felt most strongly in the external sector where import and export prices for the first half o f 1973 have only just, within the past month, become available, at a time when we could reasonably expect the annual figures. In an appendix to the Quarterly o f January 1973 i t was pointed out that i t is only the annual figure that has a meaning and the forecasts have to be based not on monthly figures but on die expected annual figure. In times o f very rapidly rising (or falling) trade prices die monthly trend tends to overestimate the annual change. While the price figures for the first six months are staggering, particularly those for exports, the resultant growth rate does not differ very significantly from the 7 per cent given in the December QEC. W e had clearly underestimated, in export prices, the price of manufactured goods due to a failure to take proper account o f the effect o f devaluation. Given the price rise in the export o f manufactures compared with the price rise i n the non-food items o f the CPI i t seems that exporters took foreign prices as given, in foreign currency units, and applied different pricing for goods for domestic sale.
I f this is the correct interpretation then the growth, i n volume terms, o f manufactured exports was very much less than we had thought. O f course, the failure to forecast adequately export prices was only part o f a general failure on export values. The October 1972 figure o f £ 7 1 1 million was replaced by the December 1973 figure o f £ 8 7 0 million for merchandise exports. Even within the broad categories, Industrial and Agricultural, the forecasts were incorrect. A substantial expected rise i n the volume o f exports o f cattle and beef, turned into an actual 8 per cent fall in such exports. W h y the forecast was wrong was explored in the August Commentary, and again i n die December Commentary—the latter being a slightly better attempt to explain the failure.
sales in a different light, and would have led to a different August issue o f QEC. By that issue we were having severe difficulties with the disposition o f industrial production between exports and home sales. Looked at w i t h hindsight i t seems industrial production was indeed growing by 13 per cent or so, manufactured exports increasing in volume terms very much less than we thought, but domestic consumption very much more—the latter, made possible by the rapid growth i n incomes. There was a further failure on consumer prices, due to an expectation that agricultural prices would not rise so rapidly in 1973, and that the operation o f the Second National Pay Agreement would lead to a reduction i n the rate o f increase o f the non-food component o f CPI.
It is possible to go through the components o f GNP and identify particular changes and reasons for particular changes in these components. This would obscure the general feature o f the earlier forecasts—a tendency to be excessively optimistic about prices, and assumptions w i t h regard to earnings, exports, etc. proving invalid over time. The forecasting exercise for 1973 was made more difficult given that the first forecast was prepared in 1972 at a time when i t was not clear what the outcome for 1972 was likely to be. Now i t is possible to identify the turning point in mid-1972. In October 1972 data was not available to indicate much more than the glimmering of such an upturn, and such data did not become available until well into 1973. O f course, this problem o f identifying turning points is terribly important for 1974. Although on average 1973 was a year o f very rapid growth it now seems clear that activity in the second half o f the year showed no increase over the first. Litde o f the growth in 1973 can thus expect to be carried into 1974.
There is also a natural tendency towards conservatism, given the wide ranging publicity Institute forecasts receive. By the April QEC i t seemed almost inevitable that the CPI would rise by a minimum o f 9*5 per cent. Yet a forecast o f that order would simply not have been believed.
Over and above these points we are now o f the opinion that too great a contact with others in the field—albeit informally—militates against a truly independent forecast. I t is idle to pretend that we are not influenced by contact, and this tends to give a certain uniformity to forecasts made by different bodies.
Conclusions
Finally, some reflections on the state o f the art o f short-term forecasting: i t should be stated that forecasting errors o f the type and magnitude experienced i n this country are also common in other countries (see [Polonyi, 3 ] , [Nelson, 4 ] for examples i n the British and American economies). These occur when the level o f sophistication in terms o f technique and the inputs o f human labour is so much greater than is the case in Ireland.
abroad [e.g. Walsh, 5 ] , it is not obvious that the returns to such exercises would be necessarily very large. Problems in such exercises are particularly acute in Ireland given its openness to exogenous influences and the fact that the quality and availability o f data is so much inferior than in countries such as the United Kingdom and the United States. Rather, we feel that further research in the effectiveness of other forms o f forecasting methods is needed before any large investment o f money is undertaken. In particular, we feel the possible extension o f survey methods should be considered. More important perhaps, what is needed is a change in the role performed by short-term forecasting and a much keener awareness o f the degree o f accuracy possible from the forecasts.
Its present role can be understood in the following terms. Assume there are n target variables in the economy (Xt... Xn) , n instrument variables ( Yx, . . ., Yn) , and n exogenous variables ( Z j . . ., Z J .
Let them be related in the following manner:
X=AY+BZ
where X, Y and Z are vectors of the various variables and A and B are vectors o f reduced form coefficients. The appropriate levels o f the instrument variables are then obtained from the equation
X* = AY+BZf:
where Zf is the forecasts o f the levels o f the exogenous variables and X * is the desired level o f the target variables. The actual level o f these latter variables however w i l l be given by
Xa = AY+BZa
where Xa and Za are the actual levels o f the target and exogenous variables. Therefore
Xa = X * - B Z H - J 3 Zf l
i.e. [Xa~X*) = B[Za-Zt)
Thus, the realised values o f the target variables will differ from the desired values i f forecasts o f exogenous variables differ from actual outcomes. I t is clear that i n some situations i f the forecasts are sufficiently inaccurate no change in policy would yield actual values o f the target variables nearer their desired levels than adjusting policy to misguided forecasts.
representing the long-term average growth rate o f the economy. Rates outside this range, however, reflect an economy which is operating at a sluggish level o f activity, or is growing at a rate which is unmaintainable over any significant period o f time. Clearly, the policy implications o f being on opposite sides o f this average range are very different. The error in the first estimates o f the growth rate have been as great as 2 and 3 per cent. The margin o f error in the forecasts must, at times, be even greater than this. Considerable errors were certainly present in 1968,1970 and 1973. I * is> obviously, possible that i f the economy was operating on one side o f the average range, forecasts, because o f errors, could conceivably reflect it operating on the other side. In such circumstances forecasting is operating in a destabilising manner. In particular, i t is obvious that the forecasts in the past have reflected the economy growing at an near average rate when, in fact, actual trends in the economy were well above (or well below) average growth.
Even i f perfect forecasts were available proper short-term "fine tuning" o f the economy could not take place. For this to occur proper techniques for relating changes in truly instrument variables, such as tax-rates, to the level o f overall economic activity i n the country would have to exist, i.e. the elements o f the A and B vectors would have to be known. Few or none such techniques exist in Ireland (or perhaps anywhere else). The implication o f this analysis appears to be that medium-term policies should take precedence over the short-term.
The implication for short-term commentaries on the economy seems to be that less emphasis should be placed on precise quantification o f future trends i n the economy, though, o f course, tables must be prepared to ensure consistency in this type o f analysis. I t should be stressed that the forecasts can, at least, indicate the general order o f magnitude o f movements i n the economy and that reasonably wide confidence intervals should be attached to any figures produced. In these circumstances less attention should be paid to mechanical manipulation o f figures and more emphasis placed on qualitative analysis o f trends in the economy.
REFERENCES
[ i ] F. T. Denton, and J. Kuiper, "The Effect of Measurement Errors on Parameter Estimates and Forecasts: A Case Study Based on Canadian Preliminary National Accounts", The Review of Economics and Statistics, May 1965.
[2] F. T. Denton, and E. H . Oksanen, "Data Revisions and Forecasting Accuracy: An
Econometric Analysis Based on Preliminary and Revised National Accounting Estimates", The Review of Income and Wealth, December 1973.
[3] G. Polonyi, "Short-Term Forecasting: A Case Study", Institute of Economic Affairs,
1973-[4] C. R. Nelson, "The Prediction Performance of the FRB-MIT-PENN Model of the U.S. Economy", The American Economic Review, December 1972.
[5] B. M . Walsh, "Econometric Macro-Model Building in the Irish Context", Quarterly
Economic Commentary, June 1970.
ABBREVIATIONS USED
QEC Quarterly Economic Commentary C Consumption G Government expenditure X Exports 1
FO First official estimate (1) CSI CSO first estimate (2) CSL CSO latest estimate (3) R -Multiple regression statistic
t ( B = i ) Computed t values under the hypothesis B = 1
Investment Imports
jth estimate in QEC Department of Finance
Durban Watson statistic Significant at 1 per cent level Significant at 5 per cent level I
M
Fi
DF D W **
*
(1) T a k e n from the annual " R e v i e w and Oudook"—prepared mainly b y the Department o f Finance.
